Les études d'association sur un génome complet (GWAS) sont conçues pour découvrir les combinaisons de points de polymorphisme (SNP) associées à des maladies. La découverte de ces associations permet d'élaborer de meilleures stratégies pour détecter, traiter ou prévenir les maladies. Récemment, l'utilisation de techniques d'extraction de patterns discriminatif a été investiguée dans le cadre de problématiques GWAS. Toutefois, la découverte de combinaisons de SNP dans de grands jeux de données GWAS est encore difficile à cause de la complexité des algorithmes utilisés. La thèse se propose donc d'améliorer l'état de l'art des approches d'extraction de motifs discriminants, dans le cadre d'extraction de combinaisons de SNP corrélées à un phénotype d'intérêt. Plusieurs solutions ont été proposées, s'attaquant aux problèmes majeurs en GWAS : évaluation de la force d'association, découverte efficace de combinaisons de SNP et visualisation de ces combinaisons. Les approches proposées sont également prometteuses pour d'autres tâches de bioinformatique comme la découverte d'expressions génique, la détection de motifs de phosphorylation et la détection de motifs de régulation. / Discovering high-order SNP combinations associated with diseases is an important task of bioinformatics. Once new genetic associations are identified, they can be used to develop better trategies to detect, treat and prevent the diseases. Recently, this issue has been effectively tackled with discriminative pattern mining algorithms. However, the number of SNPs is often very large, discovering of SNP combinations remains many challenges. To address these challenges this thesis has been advanced the state-of-the-art discriminative pattern mining techniques to discover SNP combinations associated with interesting phenotype. Different solutions have been proposed in this thesis to tackle GWAS analysis. These solutions focus on efficient association strength evaluation, statistically significant discriminative SNP combinations discovery and interesting SNP combinations visualization. The solutions proposed in this thesis are also promising for other tasks of bioinformatics such as differential gene expression discovery, phosphorylation motifs detection and regulatory motif combination mining.

4 Abstract The prevalence of obesity and its related cardiometabolic complications in children remains high across the world. Obesity is a multifactorial disease caused by interaction between genes and environmental factors. Genome-wide association studies have discovered several single nucleotide polymorphisms associated with obesity. A causal role of infection in the pathogenesis of obesity has also been considered, particularly the role of adenovirus 36 (Adv36). The aim of the Ph.D. thesis was to investigate the associations of obesity susceptibility loci (TMEM18, SH2B1, KCTD15, PCSK1, BDNF, SEC16B, MC4R, FTO) and Adv36 infection with obesity-related characteristics and complications in the Czech adolescent population. The results are described in eight publications, of which six are original papers and two are reviews. Studies were performed on a cohort of Czech adolescents recruited either from the general population (1,533 individuals from the epidemiological study) and from in-patient or outpatient weight management clinics (562 overweight/obese individuals underwent an intervention). The results demonstrated an association of TMEM18, SEC16B and FTO gene variants with obesity. Some variants of the genes involved in hypothalamic regulation of energy homeostasis − MC4R, BDNF, PCSK1 − were related to...

Males and age group 1 to 5 years show a much higher risk for childhood acute lymphoblastic leukemia (ALL). We performed a case-only genome-wide association study (GWAS), using the Illumina Infinium HumanCoreExome Chip, to unmask gender- and age-specific risk variants in 240 non-Hispanic white children with ALL recruited at Texas Children’s Cancer Center, Houston, Texas. Besides statistically most significant results, we also considered results that yielded the highest effect sizes. Existing experimental data and bioinformatic predictions were used to complement results, and to examine the biological significance of statistical results. Our study identified novel risk variants for childhood ALL. The SNP, rs4813720 (RASSF2), showed the statistically most significant gender-specific associations (P < 2 x 10-6). Likewise, rs10505918 (SOX5) yielded the lowest P value (P < 1 x 10-5) for age-specific associations, and also showed the statistically most significant association with age-at-onset (P < 1 x 10-4). Two SNPs, rs12722042 and 12722039, from the HLA-DQA1 region yielded the highest effect sizes (odds ratio (OR) = 15.7; P = 0.002) for gender-specific results, and the SNP, rs17109582 (OR = 12.5; P = 0.006), showed the highest effect size for age-specific results. Sex chromosome variants did not appear to be involved in gender-specific associations. The HLA-DQA1 SNPs belong to DQA1*01:07and confirmed previously reported male-specific association with DQA1*01:07. Twenty one of the SNPs identified as risk markers for gender- or age-specific associations were located in the transcription factor binding sites and 56 SNPs were non-synonymous variants, likely to alter protein function. Although bioinformatic analysis did not implicate a particular mechanism for gender- and age-specific associations, RASSF2 has an estrogen receptor-alpha binding site in its promoter. The unknown mechanisms may be due to lack of interest in gender- and age-specificity in associations. These results provide a foundation for further studies to examine the gender- and age-differential in childhood ALL risk. Following replication and mechanistic studies, risk factors for one gender or age group may have a potential to be used as biomarkers for targeted intervention for prevention and maybe also for treatment.

La sélection génomique substitue à la connaissance de la généalogie celle des séquences d’ADN et connait un succès spectaculaire dans la sélection des bovins laitiers. En équin, le gain de précision pour les valeurs génétiques en CSO a été estimé faible entre la généalogie et la génomique, éventuellement à cause des particularités des populations d’apprentissage et de validation. L’objectif est de définir pour les races équines les conditions d’efficacité et de fonctionnement de la sélection génomique. La partie théorique de la thèse a consisté en une méta-analyse afin de comprendre le lien entre précision théorique et observée en fonction des paramètres des populations. L’étude a montré l’importance du nombre efficace de marqueurs Me. Ce paramètre spécifique de la population, de la structure génomique et de la parenté doit être évalué, au même titre que l’héritabilité en génétique classique. D’un point de vue pratique, la 1ère voie d’amélioration était de rechercher des gènes à effet majeur sur l’aptitude au concours de saut d’obstacles (CSO) ou au concours complet. Aucun gène majeur n’a été localisé malgré des détections significatives. Le 2nd levier pour améliorer l’estimation des valeurs génétiques en CSO était d’utiliser le Single-Step, méthode qui combine l’information génomique des étalons génotypés et la généalogie de l’ensemble des chevaux non génotypés utilisés pour l’indexation. L’évaluation pour le CSO a donc été revisitée. Malgré le re-calcul de l’héritabilité et l’application des points sur toute la période, le gain en précision reste faible. La sélection génomique a également été testée sur des chevaux d’endurance, mais comme pour le CSO les précisions obtenues pour le moment ne sont pas assez élevées pour justifier une utilisation de la sélection génomique. Récemment, un gène majeur agissant sur l’aptitude à trotter (DMRT3) a été identifié. Malgré l’effet très négatif d’un allèle sur la qualification et les performances précoces, le Trotteur français (TF) est polymorphe pour le gène à cause d’un effet positif de ce même allèle sur les performances tardives. La sélection classique et la sélection génomique ont été comparées en incluant ou non dans le modèle un marqueur lié à DMRT3, nous permettant d’identifier la meilleure combinaison de modèle et de méthode à utiliser pour estimer les valeurs génétiques du TF. Enfin, le paramètre Me a été estimé dans les populations de chevaux utilisées au cours de la thèse, et les résultats des évaluations génomiques ont été comparés en fonction de Me et des autres paramètres influant sur la précision de la sélection génomique. Deux nouveaux projets prévoyant de génotyper des chevaux de CSO d’une part et des TF d’autre part devraient permettre respectivement d’améliorer la précision de l’évaluation génomique en CSO et de confirmer l’intérêt de la prise en compte de DMRT3 dans l’évaluation génomique des TF. / Genomic selection uses genotypes information instead of pedigree information for the estimation of breeding values. In dairy cattle, the selection schemes were greatly improved with this method. In horses, a first attempt of genomic selection showed that the evaluation accuracy was not much improved when using genotypes information compared to classic evaluation, possibly because of the structure of the reference and validation populations. The objective of the thesis was to define the theoretical and practical conditions for the use of genomic selection in horses. The theoretical work of the thesis consisted in a meta-analysis to understand the relation between observed and theoretical accuracy depending on the parameters of the population. We proved the importance of the effective number of independent segments in the genome Me. This parameter is specific of the population and of the genomic structure and relationship structure. We recommend to estimate this parameter before genomic evaluation, just like heritability that is estimated before genetic evaluation. Regarding practical tasks of the thesis, the first solution to improve the breeding values estimation for jumping performances was to look for genes having a major effect on performances in jumping competitions and three-day’s events, but no major gene was evidence in spite of significant detections. The 2nd solution was to perform a single-step evaluation. This method combines information from genotyped stallions and from the pedigree of the whole population. Even if the heritability was re-estimated and points distributed to all horses to have a homogeneous criteria, the accuracy of genomic evaluation was not much improved. Genomic selection was also tested on horses running endurance races, but as for jumping the accuracy was not high enough. Recently, a major gene having a huge effect on the ability of horses to trot was evidenced (DMRT3). Even if one allele has a negative effect on qualification and early earnings, French Trotter (FT) is still heterozygote because of a positive effect of this allele on late performances. Genetic and genomic evaluations were compared with or without using in the model a SNP linked to DMRT3 as a fixed effect. This study allowed identifying the best combination of model and method to use for estimation of FT breeding values. Finally, the parameter Me was estimated in the populations of horses used in the thesis. The results of genomic evaluations were compared according to Me and the other parameters having an influence on the accuracy of genomic evaluations. Two new projects will genotype more jumping horses and FT, they should allow to improve the accuracy of genomic evaluation for jumping horses and to acknowledge the interest of using DMRT3 in the genomic evaluation of FT.

Over the last ten years, genome-wide association studies (GWAS) have been used to identify genetic variants associated with many diseases as well as quantitative phenotypes, by exploiting naturally occurring genetic variation in large cohorts of individuals. More recently, the GWAS approach has also been applied to highthroughput RNA sequencing (RNA-seq) data in order to find loci associated with different levels of gene expression, called expression quantitative trait loci (eQTL). Because of the large amount of data that is required for such high-resolution eQTL studies, most of them have so far been carried out in humans, where the cost of data collection could be justified by a possible future impact in human health. However, due to the rapidly falling price of high-throughput sequencing it is now also becoming feasible to perform high-resolution eQTL studies in higher model organisms. This enables the study of gene regulation in biological contexts that have so far been beyond our reach for practical or ethical reasons, such as early embryonic development. Taking advantage of these new possibilities, we performed a high-resolution eQTL study on 80 inbred fruit fly lines from the Drosophila Genetic Reference Panel, which represent naturally occurring genetic variation in a wild population of Drosophila melanogaster. Using a 3′ Tag RNA-sequencing protocol we were able to estimate the level of expression both of genes as well as of different 3′ isoforms of the same gene. We estimated these expression levels for each line at three different stages of embryonic development, allowing us to not only improve our understanding of D. melanogaster gene regulation in general, but also investigate how gene regulation changes during development. In this thesis, I describe the processing of 3′ Tag-Seq data into both 3′ isoform expression levels and overall gene expression levels. Using these expression levels I call proximal eQTLs both common and specific to a single developmental stage with a multivariate linear mixed model approach while accounting for various confounding factors. I then investigate the properties of these eQTLs, such as their location or the gene categories enriched or depleted in eQTLs. Finally, I extend the proximal eQTL calling approach to distal variants to find gene regulatory mechanisms acting in trans. Taken together, this thesis describes the design, challenges and results of performing a multivariate eQTL study in a higher model organism and provides new insights into gene regulation in D. melanogaster during embryonic development.

L'ambition de cette thèse était de proposer un nouvel élan pour la création variétale chez la vigne, incluant les connaissances et les derniers outils de la recherche. En effet, la viticulture française comme d'autres filières agricoles doit aujourd'hui faire face à 3 grands défis: la réduction des intrants phytosanitaires (plan Ecophyto 2018), les changements climatiques, et de nouveaux concurrents, notamment les pays du Nouveau Monde. La création variétale, qui a été peu exploitée chez la vigne, peut être une des solutions pour répondre à ces défis.S'appuyant sur un génotypage dense, plusieurs outils et concepts innovants – réunis sous le terme de sélection génomique (GS) – ont vu le jour ces dernières années en sélection animale, qui permettent de prédire les phénotypes des individus seulement génotypés.Afin d'atteindre notre but, nous avons évalué et comparé l'efficacité de la GS et de la sélection assistée par marqueur (SAM) « classique », basée sur la génétique d'association « genome-wide » (GWAS) chez la vigne. Le potentiel théorique des deux méthodes a été évalué dans une étude de simulation, puis sur des données réelles.Nous montrons que la GS est plus pertinente que la SAM « classique » pour prédire les phénotypes et ce pour des caractères complexes et / ou structurés. Cependant la GS couplée aux résultats issus de la GWAS semble être une méthode intéressante lorsque le marquage moléculaire est non limitant. Finalement, nous discutons des conditions d'utilisation de la GS en termes économiques et d'efficacité au cours du temps. Nous proposons trois scénarios fonction de l'investissement de départ et des besoins en termes de création variétale. / The aim of this PhD project was to provide a new impulse for grapevine breeding, applying the latest knowledge and research tools on this species. Indeed, French viticulture, as well as various other agricultural sectors, faces today three major challenges: how to reduce phytosanitary inputs (Ecophyto 2018), impact of climatic changes and new competitors on the market, especially New World countries. Plant breeding in grapevine has not been much exploited until today, but could be a solution to these challenges.Several innovative tools and concepts have seen the light in animal breeding in the last decade. Using high density genome-wide marker information and advanced statistical models, phenotypes can be predicted for individuals that were merely genotyped. The method termed genomic selection (GS) is implementing this type of approach.To achieve our aim, we evaluated and compared the efficiency of GS and “classical” marker-assisted selection (MAS), based on genome-wide association study (GWAS) for grapevine. The theoretical potential of the two methods was evaluated in a simulation study but also on real data.We show that GS is more relevant than “classical” MAS to predict phenotypes of complex and / or structured traits. However, the combination of GS with results from GWAS seems to be of particular interest if the number of molecular markers available is adequate. Finally, we discuss GS implementation in terms of economic aspects and efficiency over time; we propose three scenarios differing by the initial investment required and the breeding objectives to be reached.

<div>This dissertation consists of two parts. The first part is a collaborative project with Dr. Szymanski's group in Agronomy at Purdue, to predict protein complex assemblies and interactions. Proteins in the leaf cytosol of Arabidopsis were fractionated using Size Exclusion Chromatography (SEC) and mixed-bed Ion Exchange Chromatography (IEX).</div><div>Protein mass spectrometry data were obtained for the two platforms of separation and two replicates of each. We combine the four data sets and conduct a series of statistical learning, including 1) data filtering, 2) a two-round hierarchical clustering to integrate multiple data types, 3) validation of clustering based on known protein complexes,</div><div>4) mining dendrogram trees for prediction of protein complexes. Our method is developed for integrative analysis of different data types and it eliminates the difficulty of choosing an appropriate cluster number in clustering analysis. It provides a statistical learning tool to globally analyze the oligomerization state of a system of protein complexes.</div><div><br></div><div><br></div><div>The second part examines global hypothesis testing under sparse alternatives and arbitrarily strong dependence. Global tests are used to aggregate information and reduce the burden of multiple testing. A common situation in modern data analysis is that variables with nonzero effects are sparse. The minimum p-value and higher criticism tests are particularly effective and more powerful than the F test under sparse alternatives. This is the common setting in genome-wide association study (GWAS) data. However, arbitrarily strong dependence among variables poses a great challenge towards the p-value calculation of these optimal tests. We develop a latent variable adjusted method to correct minimum p-value test. After adjustment, test statistics become weakly dependent and the corresponding null distributions are valid. We show that if the latent variable is not related to the response variable, power can be improved. Simulation studies show that our method is more powerful than other methods in highly sparse signal and correlated marginal tests setting. We also show its application in a real dataset.</div>

京都大学 / 新制・論文博士 / 博士(医学) / 乙第13398号 / 論医博第2222号 / 新制||医||1051(附属図書館) / (主査)教授 妹尾 浩, 教授 中山 健夫, 教授 西浦 博 / 学位規則第4条第2項該当 / Doctor of Medical Science / Kyoto University / DFAM

Les gliomes constituent les plus fréquentes des tumeurs malignes primaires du système nerveux central. Les liens qui existent entre ces tumeurs et un certain nombre de cancers rares héréditaires, comme les Neurofibromatoses I et II ou les syndromes de Turcot et de Li-Fraumeni, attestent d’une prédisposition génétique aux gliomes. L’observation d’un risque deux fois plus élevé de développer un gliome chez les parents de premier degré de patients atteints suggère aussi une possible prédisposition génétique dans les gliomes sporadiques. Par ailleurs, l’analyse à haut débit permet de préciser le profil somatique des gliomes et d’identifier des biomarqueurs pronostiques voire prédictifs et s’inscrire dans une démarche de traitement personnalisé du patient. Durant ma thèse, je me suis focalisé sur deux axes de recherches complémentaires; l’identification de gènes de susceptibilité et la découverte de nouveaux gènes fréquemment mutés dans les gliomes, afin de déterminer les voies de signalisation contribuant à la gliomagenèse. Dans leur ensemble, les résultats obtenus dans cette thèse apportent non seulement des informations importantes sur la nature de la prédisposition génétique aux gliomes mais également de son association spécifique pour les différents sous-types de tumeurs. La découverte d’un nouveau gène muté, offre la perspective à plus long terme d’un traitement personnalisé pour chaque patient sur la base du profil génétique de sa tumeur. / Gliomas are the most common adult malignant primary tumour of the central nervous system. Thus far, no environmental exposures has been linked to risk except for ionizing radiation, which only accounts for a very small number of cases. Direct evidence for inherited predisposition to glioma is provided by a number of rare inherited cancer syndromes, such as Turcot's and Li–Fraumeni syndromes, and neurofibromatosis. Even collectively, these diseases however account for little of the twofold increased risk of glioma seen in first-degree relatives of glioma patients. My research was centred on two complementary research activities: Identifying susceptibility genes for glioma to delineate key biological pathways contributing to disease pathogenesis and to identify new recurrent mutated genes for glioma to provide for further insights into glial oncogenesis and suggesting targets for novel therapeutic strategies. Collectively the findings in this thesis provide increased insight into the nature of genetic predisposition to glioma and substantiate the often distinct associations between susceptibility variants and glioma molecular groups. In addition the discovery of a new mutated gene in glioma offers the potential to support drug development and advance precision medicine for this tumours.

<p>Within the last few decades, swine breeding programs have been refined to include pork quality and novel carcass traits alongside growth, feed efficiency, and carcass leanness in the selection programs for terminal sire lines with a goal to produce high quality and efficient pork product for consumers. In order to accurately select for multiple traits at once, it becomes imperative to explore their genetic and biological architecture. The genetic architecture of traits can be explored through the estimation of genetic parameters, genome-wide association studies (GWAS), gene networks and metabolic pathways. An alternative approach to explore the genetic and biological connection between traits is based on principal component analysis (PCA), which generates novel “pseudo-phenotypes” and biological types (biotypes). In this context, the main objective of this thesis was to understand the genetic and biological relationship between three growth, eight conventional carcass, 10 pork quality, and 18 novel carcass traits included in two studies. The phenotypic data set included 2,583 records from female Duroc pigs from a terminal sire line. The pedigree file contained 193,764 animals and the genotype file included 21,344 animals with 35,651 single nucleotide polymorphisms (SNPs). The results of the first study indicate that genetic progress can be achieved for all 39 traits. In general, the heritability estimates were moderate, while most genetic correlations were generally moderate to high and favorable. Some antagonisms were observed but those genetic correlations were low to moderate in nature. Thus, these relationships can be considered when developing selection indexes. The second study showed that there are strong links between traits through their principal components (PCs). The main PCs identified are linked to biotypes related to growth, muscle and fat deposition, pork color, and body composition. The PCs were also used as pseudo-phenotypes in the GWAS analysis, which identified important candidate genes and metabolic pathways linked to each biotype. All of this evidence links valuable variables such as belly, color, marbling, and leanness traits. Our findings greatly contribute to the optimization of genetic and genomic selection for the inclusion of valuable and novel traits to improve productive efficiency, novel carcass, and meat quality traits in terminal sire lines.<br></p><p></p>